KR102161905B1 - Backward vehicle detection apparatus and method - Google Patents

Abstract

The present invention discloses a rear vehicle detection apparatus and method. According to the present invention, there is provided a processor; And a memory connected to the processor, wherein the memory determines a vanishing point through recognition of a driving lane in the rear image input through the camera, and sets a first region of interest of a preset size below the determined vanishing point. And storing program instructions executable by the processor to determine whether a shadow exists by searching for a shadow candidate region in the first region of interest, and performing edge pairing on the searched shadow candidate region. A vehicle detection device is provided.

Description

Backward vehicle detection apparatus and method TECHNICAL FIELD

The present invention relates to an apparatus and method for detecting a rear vehicle, and more particularly, to an apparatus and method capable of preventing a collision with a rear vehicle during high-speed driving.

Traffic accidents frequently occur due to the driver's carelessness, poor visibility, and violation of the duty to look forward to the vehicle behind the vehicle.

To prevent this, vehicles are recently equipped with Advanced Driver Assistance Systems (ADAS).

The Intelligent Driver Assistance System uses advanced detection sensors, GPS, communication, and intelligent video equipment to recognize some situations while driving by itself to determine the situation, control the car or detect dangers in advance by sound and light. It is a driver assistance system that informs you by vibration, etc.

Among them, the Backward Collision Warning System (BCWS) is a system that determines whether there is a vehicle in the rear through a camera, and provides a warning to the driver if it is located within a certain distance.

Conventionally, the rear vehicle collision avoidance system detects the contour of an object existing at the rear, which has a high complexity problem.

Korean Patent Registration 10-1478072

In order to solve the above-described problems of the prior art, an apparatus and method for detecting a rear vehicle capable of preventing a rear vehicle collision during high-speed driving while reducing the complexity are proposed.

In order to achieve the object as described above, according to an embodiment of the present invention, a processor; And a memory connected to the processor, wherein the memory determines a vanishing point through recognition of a driving lane in the rear image input through the camera, and sets a first region of interest of a preset size below the determined vanishing point. And storing program instructions executable by the processor to determine whether a shadow exists by searching for a shadow candidate region in the first region of interest, and performing edge pairing on the searched shadow candidate region. A vehicle detection device is provided.

The program instructions calculate an average road brightness value by using pixels in a region adjacent to the first region of interest to search for the shadow candidate region, and divide the first region of interest into a plurality of sections in a height direction to each section. The brightness value of is calculated, and it may be determined whether the first section having the minimum brightness value is located in the center of the traveling lane.

The program instructions may set a second ROI adjacent to the shadow candidate area for edge pairing, and determine whether an edge symmetrical to the second ROI exists.

The program instructions generate a y-axis edge map of the second ROI, and an edge magnitude of the edge map to determine whether an edge symmetrical exists in the second ROI. A point greater than or equal to a preset value may be determined as the edges, and it may be determined whether the edges are symmetrical to each other through horizontal inversion based on the center point of the edges.

The program instructions may count an accumulation number of frames in which a shadow exists in a predetermined area of the traveling lane, and generate an alarm when the accumulation number is greater than or equal to a preset number.

After generating the alarm, the program instructions may determine whether the position of the shadow is closer than the position at which the alarm is generated, and regenerate the alarm when the distance to the shadow becomes close.

The program instructions may detect the rear vehicle through whether or not pixels having a predetermined brightness or more symmetrically exist in one or more regions of interest in order to detect the vehicle behind the vehicle in a night situation.

The program instructions, when a pixel having a predetermined brightness or more symmetrically exist in the one or more regions of interest, set the pixel having a predetermined brightness or higher as a headlight candidate region, and are preset in the upper portion of the preset size of the headlight candidate region. It may be determined whether or not the headlight candidate region is the final headlight based on whether or not there are pixels of brightness or higher.

The program commands may determine the y value of the first ROI by using a distance d1 between a rear vehicle and a user vehicle, and a ratio of a distance d2 between a vanishing point and a bottom of the rear vehicle.

The program commands estimate the camera tilt or height using the y value of the vanishing point in the image, and determine the graph trend lines of d1 and d2 in consideration of the estimated tilt or height to determine the y value of the first region of interest. I can.

According to another aspect of the present invention, there is provided an apparatus for detecting a rear vehicle, comprising: a processor; And

Including a memory connected to the processor, wherein the memory determines a vanishing point through recognition of a driving lane in the rear image input through the camera, and sets a first region of interest of a preset size at a lower end of the determined vanishing point, and , Using a plurality of headlight search boxes of different sizes in the first region of interest, search for a candidate headlight region, and a predetermined region above the first headlight search box corresponding to the candidate headlight region as a first window search box. A rear vehicle detection device storing program instructions executable by the processor to determine that a headlight exists in the first headlight search box when the brightness in the first window search box is lower than a preset threshold Is provided.

According to another aspect of the present invention, there is provided a method for detecting a rear vehicle, the method comprising: determining a vanishing point through recognition of a driving lane from a rear image input through a camera; Setting a first region of interest of a predetermined size below the determined vanishing point; Searching for a shadow candidate area within the first ROI; There is provided a rear vehicle detection method comprising the step of determining whether a shadow exists by performing edge pairing on the searched shadow candidate region.

According to the present invention, since the rear vehicle is detected through shadow search and edge pairing in the region of interest, it is possible to detect the rear vehicle with low complexity.

In addition, when driving at night, there is an advantage of being able to detect a rear vehicle through headlight search and headlight pairing in a region of interest.

1 is a view showing the configuration of a rear vehicle detection apparatus according to an embodiment of the present invention.
2 is a diagram illustrating a module of program instructions for detecting a rear vehicle according to the present embodiment.
3 is a diagram illustrating a process of determining a vanishing point according to the present embodiment.
4 is a diagram illustrating a first region of interest and an adjacent region for calculating an average road brightness value according to the present embodiment.
5 is a diagram illustrating a horizontal area for determining a shadow candidate area according to the present embodiment.
6 is a diagram illustrating a second ROI for edge pairing according to the present embodiment.
7 is a diagram for explaining a headlight search process for detecting a rear vehicle according to the present embodiment.
FIG. 8 is a diagram illustrating a case in which the headlights of young vehicles located at the rear due to the bright headlights are mistaken for headlights by one vehicle.
9 is a view showing a horizon according to an embodiment of the present invention.
10 is a diagram illustrating a region of interest in a night situation according to an embodiment of the present invention.
11 is a diagram illustrating a headlight search box according to the present embodiment.
12 shows an area in which misrecognition can occur according to light reflected on a road surface.
13 illustrates an area in which misrecognition may occur due to a vehicle located in another lane.
14 is a diagram illustrating a window search box according to an embodiment of the present invention.

In the present invention, various modifications may be made and various embodiments may be provided, and specific embodiments will be illustrated in the drawings and described in detail.

However, this is not intended to limit the present invention to a specific embodiment, it is to be understood to include all changes, equivalents, and substitutes included in the spirit and scope of the present invention.

1 is a view showing the configuration of a rear vehicle detection apparatus according to an embodiment of the present invention.

As shown in FIG. 1, the apparatus for detecting a rear vehicle according to the present embodiment may include a processor 100 and a memory 102.

The processor 100 may include a central processing unit (CPU) capable of executing a computer program or a virtual machine.

The memory 102 may include a nonvolatile storage device such as a fixed hard drive or a removable storage device. The removable storage device may include a compact flash unit, a USB memory stick, or the like. The memory 202 may also include volatile memories such as various random access memories.

Program instructions executable by the processor 100 are stored in the memory 102.

The rear vehicle detection apparatus according to the present embodiment is a computing device installed in a vehicle, and detects a rear vehicle by analyzing an image input through a camera and generates an alarm.

Here, the camera may be a camera installed in a black box, but is not limited thereto.

According to a preferred embodiment of the present invention, the memory 102 stores program instructions for recognizing a lane in an image input through a camera, determining a vanishing point, and determining whether a vehicle is present in the region of interest.

According to an embodiment of the present invention, when the vehicle is driving at high speed, the rear vehicle detection apparatus assumes that the object existing in a predetermined area of the rear image for a certain period of time is a vehicle. The vehicle is detected.

The rear vehicle detection device searches for a shadow candidate region through comparison of the brightness values of pixels, and determines whether the shadow candidate region is a shadow by a vehicle located behind the driving lane of the user vehicle through an edge pairing process. .

The rear vehicle detection apparatus counts the accumulated number of frames in which a shadow exists, and generates an alarm when the number of shadow accumulation exceeds a threshold.

Since the search for the shadow is performed based on a predetermined region of interest, the presence of the shadow in the region of interest means that it is within a predetermined distance, and the rear vehicle detection apparatus generates an alarm.

To prevent redundant alarms, an alarm is additionally generated when the distance to the shadow becomes close after one alarm occurs.

On the other hand, it can be difficult to search for shadows at night.

In consideration of this, the rear vehicle is detected by searching the headlights of the rear vehicle at night.

Hereinafter, a process of detecting a rear vehicle according to the present embodiment will be described in detail with reference to the drawings.

2 is a diagram illustrating a module of program instructions for detecting a rear vehicle according to the present embodiment.

FIG. 2 shows a module functionally classifying program instructions stored in the memory 102 for detection of a rear vehicle.

Referring to FIG. 2, the module according to the present embodiment may include a vanishing point determination module 200, an ROI setting module 202, a shadow search module 204, and an alarm generation module 206.

As shown in FIG. 3, the vanishing point determination module 200 recognizes both lanes behind the driving direction and determines the intersection of the recognized driving lanes as the vanishing point.

The region of interest setting module 202 sets a preset lower region of the vanishing point as a first region of interest.

As shown in FIG. 4, the first ROI 400 is set in a rectangular shape in a lower area of the vanishing point. The first region of interest 400 is not limited to a rectangular shape and may have various shapes.

According to an embodiment of the present invention, the y value (y position) of the first ROI 400 may be determined in the following manner. Here, the y value of the first ROI 400 may be a coordinate on the y-axis of the upper end of the first ROI 400 on the image frame.

Here, in the image of FIG. 4, the horizontal direction is defined as the x-axis and the vertical direction is the y-axis, and the y value may be defined as the highest value of the y-axis of the first ROI 400.

In actual road conditions, when the distance between the rear vehicle and the user's vehicle is d1, and the distance between the vanishing point and the rear surface of the vehicle in the image is d2, the smaller d1, the larger the d2 change, and the larger d1, the d2 The change in can be small.

Considering this point, a graph trend line of d1:d2 is determined, and when d1 for a rear vehicle alarm is determined, a y value of the first ROI is determined in consideration of d2 according to the determined d1.

Meanwhile, the graph trend line of d1:d2 may vary depending on the camera tilt.

When the camera tilt has a negative value, that is, when it faces backward and downward, the y value of the vanishing point may increase, and conversely, when the camera tilt has a positive value, the y value of the vanishing point may decrease.

When rearviewing is started in consideration of this point, when the y value of the vanishing point is determined, a camera tilt is inferred through this, and a d1:d2 graph trend line is determined.

The determination of the graph trend line may vary depending on the camera height as well as the camera tilt.

The shadow search module 204 searches for a shadow by a vehicle located behind the traveling lane in the first ROI 400.

The shadow search module 204 calculates an average road brightness value Road _avg using pixels in an area adjacent to the first ROI 400 of FIG. 4.

In addition, the first ROI 400 is divided into a plurality of sections (eg, 5 pixels) in the height direction (y-axis direction) to calculate the brightness value of each section, and the minimum brightness value among the plurality of sections is calculated. The branch searches for the first section.

Here, a shadow candidate region may exist in the first section having the minimum brightness value.

The shadow search module 204 estimates that a shadow candidate region exists when the brightness value of the first section having the minimum brightness value is less than or equal to a preset ratio of the average brightness value of the road (eg, less than 80%).

In order to determine whether the shadow candidate area is a shadow caused by a rear vehicle, the shadow search module 204 determines whether the first section is located in the center of the traveling lane.

As shown in FIG. 5, the shadow search module 204 generates an area (horizontal area, 500) that includes a first section and extends in the horizontal direction, and divides the horizontal area into a plurality of sections in the width direction, Calculate the brightness value of the section.

Here, the length in the width direction of the horizontal region 500 may be set to a length including at least a part of the left and right lanes based on the vanishing point.

The shadow search module 204 determines whether the first section is located in the center of the horizontal area 500.

When the first section is not located in the center of the horizontal area 500, the corresponding shadow may be caused by a vehicle located in a lane next to the driving lane.

When the first section is located in the center of the horizontal area 500, the shadow search module 204 determines the section as a shadow candidate area and performs an edge pairing process.

For edge pairing, the ROI setting module 202 generates a second ROI having a preset size in contact with the shadow candidate area.

The shadow search module 204 performs edge pairing within the second ROI.

6 is a diagram for explaining a process of edge pairing according to the present embodiment.

The shadow search module 204 generates a y-axis edge map of the second ROI.

It is determined whether the edges 600 and 602 having the largest values are symmetrical to each other based on the edge magnitude of the edge map.

According to the present embodiment, for edge pairing, template matching is performed for comparing edge maps with respect to a central point in the x-axis direction of edges that are symmetrical to each other.

In template matching, one of the left and right sections is reversed and compared.

When it is confirmed that the edges are symmetrical to each other in template matching, the shadow search module 204 determines that the shadow candidate region is a shadow by a vehicle located behind the traveling lane.

The alarm generation module 206 counts the accumulated number of frames in which a shadow exists, and generates an alarm when the number of accumulated shadows exceeds a threshold value.

After the alarm is generated, in order to prevent the alarm from being repeatedly generated by the rear vehicle located at the same distance, the alarm generation module 206 is used only when the position of the shadow is closer to the position of the user vehicle than the position at the time of generating the alarm. Generate an alarm.

According to the present invention, in generating an alarm, in consideration of the fact that the object remaining in the rear during high-speed driving is a vehicle and that the edge symmetrical to the rear is a part of the vehicle, the detection of the shadow without detecting the outline of the detected object You can only create an alarm.

At night, it may be difficult to search for a shadow as in the daytime, and according to the present embodiment, the search for a shadow proceeds to search for a headlight.

7 is a diagram for explaining a headlight search process for detecting a rear vehicle according to the present embodiment.

Referring to FIG. 7, if there are more than a predetermined number of pixels having a predetermined brightness or higher in an ROI, it is regarded as a headlight.

Consider that the position of the headlights is generally symmetric.

According to the present embodiment, the region of interest includes a first region of interest below the frame and a second region of interest above the frame, and the rear vehicle detection apparatus determines whether a headlight exists in each region of interest.

The rear vehicle detection apparatus searches for a candidate headlight area from the left side from the center of each of the first and second ROI.

When a pixel having a brightness greater than or equal to a preset brightness is searched on the left, the candidate area of the headlight is searched by moving a predetermined distance to the right in the x-axis direction of the same height.

If there are more than a predetermined number of pixels having a predetermined brightness or more on the right, it is determined as the left and right headlights, and counted.

However, in a night situation, the headlights of younger vehicles located in the rear may be mistaken as the headlights of one vehicle due to the bright headlights as shown in FIG. 8.

To this end, the rear vehicle detection apparatus according to the present embodiment examines the vicinity of the headlight and determines whether the searched pixel having a predetermined brightness or higher is a headlight of a rear vehicle located in a driving lane.

Since most vehicles actually do not have a device that emits light other than the headlights, it is assumed that there will be no other light around the headlights, and the upper part of the left and right headlight candidate areas is searched.

The upper part of the candidate headlight area according to the present embodiment may be set as follows.

이고, 탐색 영역 좌측 상단의 y좌표는

이다. Here, the x-coordinate of the upper left of the upper part of the candidate headlight area (the search area) is

And the y-coordinate at the top left of the search area is

to be.

은 상기한 관심 영역에서 검출된 미리 설정된 밝기 이상을 갖는 픽셀들의 최소 y좌표이고,

는 영상의 너비이다. In addition,

Is the minimum y-coordinate of pixels having more than a preset brightness detected in the region of interest,

Is the width of the image.

The apparatus for detecting a rear vehicle according to the present embodiment determines whether or not a pixel having a predetermined brightness or higher is present in the search area other than the headlight.

The rear vehicle detection apparatus generates an alarm when it is determined that the headlight candidate area is caused by a vehicle in the rearward direction of the headlight candidate area through the search of the upper portion of the headlight candidate area.

According to a preferred embodiment of the present invention, in order to increase the accuracy of detection of a vehicle behind a vehicle in a night situation, a lane is used to reduce misrecognition by other lane vehicles and infer a more accurate distance.

In a night situation, as shown in FIG. 9, the rear vehicle detection apparatus according to the present embodiment detects a driving lane through a rear image and obtains horizon information 900 through this.

Here, the acquisition of the horizon information 900 may be defined as a process of searching for a vanishing point and determining a horizontal direction (x-axis direction) of the vanishing point as the horizon.

Thereafter, as shown in FIG. 10, a predetermined area under the acquired horizon is set as the ROI 1000.

)을 탐색한다.The rear vehicle detection device is a Light-Dark-Light section (e.g.,

).

와 같이 white box 밝기와 black box 밝기 차가 일정 이상인 구간을 탐색하여 전조등 후보 영역을 탐색하는 과정일 수 있다. The light-dark-light section search process

As described above, it may be a process of searching for a headlight candidate region by searching for a section in which the difference between the white box brightness and the black box brightness is more than a certain level.

11 is a diagram illustrating a headlight search box according to the present embodiment.

As shown in FIG. 11, the size and area of each of the plurality of headlight search boxes 1100 are set to increase from the horizon to the user vehicle, that is, from top to bottom in the image.

The increase rate may be set as a proportional expression of the y position difference with the horizon in consideration of the camera angle of view.

Since the size of the box may be different depending on the rear vehicle type, camera installation location, and light spread, the light-dark-light section search is repeatedly performed by increasing the scale of the headlight search box by a predetermined ratio (for example, 20 percent).

In order not to react to the light reflected on the road surface or lane, the Light-Dark-Light section where the white box brightness (average brightness) is less than a predetermined threshold is removed.

In the case of a sensor with severe gain control, it reacts to the light reflected on the road surface during the above-described light-dark-light section search process, and in the case of corners, many reactions with other lane vehicles occur.

FIG. 12 is a diagram illustrating a region in which erroneous recognition may occur due to light reflected on a road surface, and FIG. 13 is a diagram illustrating a region in which erroneous recognition may occur due to a vehicle located in another lane.

In FIGS. 12 to 13, red boxes 1200 and 1300 are areas in which erroneous recognition can occur.

In order to minimize such misrecognition, in the case of the front window area of the rear vehicle, an image characteristic in which light does not exist is used.

To this end, as shown in FIG. 14, the rear vehicle detection apparatus generates a binary image by setting the upper area as the window search box 1400 by a predetermined distance from the headlight search box including the Light-Dark-Light section.

Preferably, since the front window is located behind the headlight in the image, the window search box is set smaller than that of the headlight search box.

The binary image is an image that is set to 255 above a predetermined threshold and 0 below it.

When a Light-Dark-Light section is searched in some of the plurality of headlight search boxes of FIG. 11, the rear vehicle detection apparatus is a binary image of a window search box on the upper side of the headlight search box including the Light-Dark-Light section. Is generated to determine whether a pixel having a brightness greater than or equal to a predetermined threshold is included.

As described above, since light should not exist in the window area, when the average brightness value of the front window search box is greater than or equal to a preset threshold, it is determined that there is no headlight in the corresponding headlight search box.

Meanwhile, when the window search box is less than a predetermined brightness, the rear vehicle detection apparatus determines that a headlight exists in the corresponding headlight search box, and generates an alarm when the frame is maintained for a predetermined time.

The above-described embodiments of the present invention have been disclosed for the purpose of illustration, and those skilled in the art who have ordinary knowledge of the present invention will be able to make various modifications, changes, and additions within the spirit and scope of the present invention, and such modifications, changes and additions It should be seen as belonging to the following claims.

Claims (13)

As a rear vehicle detection device,
Processor; And
Including a memory connected to the processor,
The memory,
The vanishing point is determined through recognition of the driving lane from the rear image input through the camera,
Setting a first region of interest of a preset size at the lower end of the determined vanishing point,
Searching for a shadow candidate region within the first region of interest,
To determine whether a shadow exists by performing edge pairing on the found shadow candidate region,
Store program instructions executable by the processor,
The program instructions,
A rear vehicle detection device that determines the y value of the first region of interest by using a distance d1 between a rear vehicle and a user vehicle and a distance d2 between a vanishing point and a rear vehicle bottom. The method of claim 1,
The program instructions for searching the shadow candidate area,
An average road brightness value is calculated using pixels in a region adjacent to the first region of interest,
Dividing the first region of interest into a plurality of sections in a height direction to calculate a brightness value of each section,
A rear vehicle detection device that determines whether a first section having a minimum brightness value is located in a center of a traveling lane. The method of claim 1,
The program instructions for the edge pairing,
Setting a second region of interest adjacent to the shadow candidate region,
A rear vehicle detection device that determines whether a symmetrical edge exists in the second ROI. The method of claim 3,
The program instructions to determine whether there is a symmetrical edge in the second region of interest,
Generate a y-axis edge map of the second ROI,
A rear vehicle detection apparatus configured to determine a point where the edge magnitude of the edge map is equal to or greater than a preset value as edges, and to determine whether the edges are symmetrical to each other through left and right inversion based on the center point of the edges. The method of claim 1,
The program instructions count the cumulative number of frames in which a shadow exists in a predetermined area of the traveling lane,
A rear vehicle detection device that generates an alarm when the accumulated number is more than a preset number. The method of claim 5,
The program instructions, after generating the alarm, determine whether the position of the shadow is closer than the position at the time of generating the alarm,
A rear vehicle detection device that generates an alarm again when the distance to the shadow approaches. The method of claim 1,
The program instructions are used to detect a rear vehicle in a night situation,
A rear vehicle detection apparatus for detecting a rear vehicle through whether or not pixels having a predetermined brightness or more symmetrically exist in one or more regions of interest. The method of claim 7,
The program instructions,
When pixels with a predetermined brightness or higher are symmetrically present in the one or more regions of interest, the predetermined brightness or higher pixel is set as a headlight candidate region, and a pixel with a predetermined brightness or higher is present in an upper portion of the predetermined size of the headlight candidate region. The rear vehicle detection device determines whether the candidate headlight area is a final headlight through whether or not. delete The method of claim 1,
The program instructions,
A rear vehicle detection device that estimates the camera tilt or height using the y value of the vanishing point in the image, and determines the y value of the first region of interest by determining the graph trend lines of d1 and d2 in consideration of the estimated tilt or height . As a rear vehicle detection device,
Processor; And
Including a memory connected to the processor,
The memory,
The vanishing point is determined through recognition of the driving lane from the rear image input through the camera,
Setting a first region of interest of a preset size at the lower end of the determined vanishing point,
Searching for a headlight candidate area using a plurality of headlight search boxes of different sizes within the first ROI,
A predetermined area above the first headlight search box corresponding to the headlight candidate area is set as the first window search box,
When the brightness in the first window search box is lower than a preset threshold, determining that a headlight exists in the first headlight search box,
Store program instructions executable by the processor,
The program instructions,
A rear vehicle detection device that determines the y value of the first ROI by using a distance d1 between a rear vehicle and a user vehicle and a distance d2 between a vanishing point and a rear vehicle bottom. The method of claim 11,
The program instructions count the cumulative number of frames in which headlights exist in a predetermined area of the traveling lane,
A rear vehicle detection device that generates an alarm when the accumulated number is more than a preset number. As a rear vehicle detection method,
Determining a vanishing point through recognition of a driving lane in the rear image input through the camera;
Setting a first region of interest of a predetermined size below the determined vanishing point;
Searching for a shadow candidate area within the first ROI; And
Comprising the step of determining whether a shadow exists by performing edge pairing on the searched shadow candidate region,
The step of setting the first region of interest may include determining a y value of the first region of interest using a ratio of a distance d1 between a rear vehicle and a user vehicle and a distance d2 between a vanishing point and a bottom of the rear vehicle. Rear vehicle detection method.

Images